Materials and Methods
Inclusion criteria for this study were: >18 years of age, patients undergoing prophylactic, oncologic, unilateral or bilateral mastectomy or lumpectomy followed by breast reconstruction, and availability of a recent breast MRI scan. Exclusion criteria were breast implants, tissue expanders, or bilateral mastectomy in the past.
No breast surgery had occurred between MR image acquisition and breast volume assessment by the plastic surgeon and 3D breast measurements. Estimation of both breast volume and 3D-images are part of standard care at our institute. The study was approved by the institute's medical ethical committee. After providing written informed consent, surgical estimation values, 3D-images, and MRI-scans were acquired and collected in a central database.
A Vectra XT 3D imaging system (Canfield Imaging Systems, Fairfield, N.J.) is used as part of standard care at our hospital. 3D images are taken after every pre-operative consultation for breast reconstruction. The breasts were photographed in standing position with the arms positioned akimbo. The camera system is adjustable to the body height. It contains six color cameras positioned in a triangulated configuration. This camera system can capture images in 180 degrees. Vectra software (Vectra Breast Sculptor, v 5.5.7, Canfield Scientific Inc) automatically processes the images into a high-resolution 3D image model. These images were used for calculation of breast volumes.
To determine breast boundaries, the Vectra XT software automatically identifies landmarks on the following locations: sternal notch (SN), midclavicular, nipple, areola, medial mammary fold (MMF), and lateral mammary fold (LMF). If automatic landmark detection was unsuccessful or unsatisfactory, the landmarks were manually placed or adjusted. From these landmarks, a region of interest (ROI) is formed (see Figure 1A). The ROI was formed by a medial border, cranial border, lateral border, and caudal border. The medial border comes from a fractional line (0.7) from the midpoint of the MMF landmarks to SN. The caudal border is derived from a circle formed by the MMF, IMF, and LMF landmarks, with an extra radius of factor 1.22 for a good measure. The lateral end goes little past the LMF point—the MMF to LMF circle is divided into 12 segments and then there is 1 additional segment past the LMF. The cranial border comes from the SN to midclavicular, the length of the upper margin is a fractional line (0.8) from MMF to LMF. The lateral border is formed by the shortest path from the lateral end of the cranial border to the lateral end of the caudal border. The main challenge with trunk surface 3D scans is the delimitation of the breast's dorsal boundary called the "chest wall." The chest wall plane is a curved plane that matches the patient's torso. The internal boundary is set at the skin level and guided by the shape of the skin surface around the breast.
Three-dimensional image of the breast. A, ROI is defined by the landmarks of the breast. B, Automatic calculation of breast volume with landmarks, using Vectra Breast Sculptor.
Now, the volume is calculated from the ROI as closed object with the defined chest wall. The Vectra XT 3D imaging system has the algorithm for calculation of breast volume (a closed object of ROI and defined chest wall) using the previous described method integrated in its system. It is processed automatically and takes only a few seconds (Figure 1B).
Assessments of breast volume for 3D images were done independently by two investigators (RK and MP) of this study. Both investigators were blinded for breast MRI measurements, the surgeon's estimation of breast volume, and for the 3D analysis from the other investigator.
Breast MRI examinations were performed for assessing disease extent in invasive lobular carcinoma, response monitoring during neoadjuvant chemotherapy, or breast cancer screening in high risk patients (eg, BRCA gene mutation carriers). At our institute, a Philips Ingenia 1.5T MRI system is used. Also existing MRI-scans from other hospitals, when available, were used in case they were the most recent scans. Volumes were calculated on T2-weighted sequences, as these were considered to best represent the anatomic landmarks needed for this measurement. There were no relevant differences in the sequence protocol settings of the different systems (Table 1). One MRI image (one breast image) was excluded because there was only a T1-weighted sequence available and parameters were different from other scans. First, multiplanar reformatted images of the breast with a thickness of 3 mm was created. Then, the cranial and caudal boundaries of the breast were defined by drawing an ROI on the sagittal plane at the level of the nipple. For this ROI, the breast contour was followed using the borders of the convexity of the breast to define the lower and upper borders (Figure 2A). Next, a switch was made to the axial plane and, using the previously defined borders, new ROIs were drawn at the most cranial and caudal image. Using a similar method as in the sagittal plane, the breast contours were followed and the medial and lateral boundaries were defined by drawing the line perpendicular to the major pectoral muscle. Between these upper and lower boundaries, additional ROIs were drawn on the axial images (Figure 2B). As a rule of thumb every fourth slice was used, unless there were significant alterations in the breast contour. The software Syngo.via (Siemens Healthcare GmbH, Erlangen, Germany) could interpolate the predefined ROIs into a continuous volume of interest, computed as cubic centimeters (cm3). These results were subsequently verified and if necessary "the nudge tool" was used for small corrections. All measurements were conducted by two independent and blinded radiology residents (N. dV, C. vB), who are specialized in breast imaging.
MRI image of the breast. A, For this ROI, the breast contour was followed using the borders of the convexity of the breast to define the lower and upper borders. B, Several extra ROIs were drawn on the axial images in between the upper and lower boundary.
A total of seven plastic surgeons took part in this study. They assessed the distance from sternal notch to nipple, breast width at its widest point, and projection of the breast from the thorax by measuring tape to assist in the estimation of breast volume. Assessment was done with the patient in standing position. If possible, measurements were performed on both breasts. At last, plastic surgeons were asked for an estimation of breast volume (in cm3). They were blinded for results of breast volume measured by 3D imaging or breast MRI.
Statistical analysis was performed using SPSS (IBM SPSS Statistics, v 23.0.0, IBM Corporation). Descriptive statistics were used to describe patient characteristics. Mean (SD) values of 3D images and MRI and plastic surgeon's estimation were used. Absolute and relative differences between 3D and MRI were given.
The level of reliability of breast volume assessment by 3D and MRI was analyzed using the intracIass correlation coefficient (ICC). Inter-rater reliability was calculated for breast volume assessment by 3D and MRI, also using ICC. ICC values of 0.00–0.20, 0.21–0.40, 0.41–0.60, 0.61–0.80, and 0.81–1.00 were used to indicate poor, fair, moderate, substantial, and excellent to perfect reliability, respectively.
However, the ICC value can be high if the methods show a similar variation pattern, even if the measurement results do not indicate a high agreement between the methods. Thus, Bland–Altman plots were used to analyze the agreement between single measurements, as well as for the measurements from the separate techniques. Limits of agreement were determined using the mean difference in volumes ±1.96 SD of the volume difference. If necessary, proportional and systematic differences were determined using a linear regression analysis of the difference and the mean of the two variables. Pearson's correlation was used to evaluate the correlation between variables. Paired samples t-test was used to test for differences between the three methods (plastic surgeons' estimation, 3D, MRI). A value of P < 0.05 was considered to be statistically significant.
Plast Reconstr Surg Glob Open. 2020;8(11):e3236 © 2020 Lippincott Williams & Wilkins